Actually, if it's a question that is proposed on a previous survey, we might look at the response to that question on previous surveys. If there was a high number of blank responses to the question, it's an indicator that the employees didn't understand that question enough to answer it. If they checked off “not applicable”, when the question could have been applicable to them, that's the first clue.
Then, too, we would engage in a focus group. It'd be in a room with about as many people as there are here today, and we would propose different versions of the question on a screen. This is one technique. The focus group might say, “I think I understand this question better than question two, so I suggest you go with that first question”. That's a brief description of how the focus-group process works and how we would help determine what is a good or bad question.